US 8184744 B2
A receiver and associated method estimates a channel impulse response of the communications signal. A communications signal is received as a burst of transmitted symbols, including a known training sequence. The joint estimation of timing offset and initial channel impulse response is determined based on the cross-correlations of the known transmitted symbols and received communications signal. A constant modulus interference removal iteration is applied to improve the initial channel impulse estimation.
1. A method of processing a received communications signal, comprising:
estimating an initial Channel Impulse Response (CIR) of the received communications signal based on a timing offset obtained as cross-correlations between a training sequence and received communications signal; and
applying a constant modulus interference removal iteration to improve the initial channel impulse estimation.
2. The method according to
3. The method according to
4. The method according to
5. The method according to
6. The method according to
7. The method according to
8. The method according to
9. The method according to
10. The method according to
11. A receiver comprising:
a channel estimator circuit configured to receive a communications signal and estimate channel impulse response components based on a timing offset obtained as cross-correlations between a training sequence and received communications signal; and
an equalizer circuit configured to receive the channel impulse response components and estimate an initial Channel Impulse Response (CIR) of the received communications signal and apply a constant modulus interference removal iteration to improve the initial channel impulse estimation.
12. The receiver according to
13. The receiver according to
14. The receiver according to
15. The receiver according to
16. The receiver according to
17. The receiver according to
This application is a continuation of Ser. No. 12/686,453 filed Jan. 13, 2010 (now U.S. Pat. No. 7,970,078 issued on Jun. 28, 2011), which is a continuation of Ser. No. 11/398,043 filed Apr. 5, 2006 (now U.S. Pat. No. 7,672,412), the disclosures of which are hereby incorporated by reference in their entirety.
The present invention relates to the field of communications systems, and more particularly, to estimating the channel impulse response for reducing interference within communications systems.
Cellular networks are expanding as more users turn to mobile devices as their primary and secondary communications devices, thus requiring an increase in the amount of wireless infrastructure. The increased load demands on the infrastructure often causes network congestion. This mandates an increase to the capacity of the communications networks, such as the GSM/GPRS (General Packet Radio Service), to reduce interference. Different proposals to increase capacity and reduce interference include systems for frequency planning or quality of signal control, power control schemes, and mobile station performance enhancement systems.
Wireless and cellular networks also suffer from increased interference emanating from other users, such as neighboring cells on the same frequency, known as co-channel interference (CCI), or interference emanating from neighboring frequencies on the same cell, known as adjacent channel interference (ACI). A single-antenna interference cancellation (SAIC) system is sometimes used to counter CCI. The use of this type of system has been standardized by the 3G-partnership project (3GPP), as the technical specification group (TSG) for the GSM/EDGE radio-access network (GERAN).
SAIC can reduce interference using Downlink Advanced Receiver Performance (DARP) techniques, as adopted by the 3GPP. The 3GPP technical specification 05.05V8.11, entitled the Third Generation Partnership Project, Technical Specification Group GSM/EDGE Radio Access Network, Radio Transmission and Reception, is hereby incorporated by reference in its entirety. This specification describes the requirements for different components used in such communications systems, for example, the transceivers, base stations, and other components.
Different SAIC algorithms used as interference reduction systems have been proposed, including a constant modulus method and joint demodulation method (JDM), which uses the joint demodulation of the user and an interference sequence. These systems typically begin with a standard least-squares (LS) estimate of a propagation channel, and a static channel profile for an interferer. A joint branch metric can be minimized using modified Viterbi systems, and any estimated sequences for an interfering signal in a least mean squares (LMS) algorithm can update channel estimates for a desired interfering propagation channel.
Such systems are disclosed in U.S. Pat. No. 7,006,811, U.S. Patent Publication No. 2004/0170234, and an article by the Nokia Research Center entitled Co-Channel Interference Suppression For Constant Modulus Signals. These systems describe an iterative method for improving channel impulse response (CIR) estimation. In such systems, the timing offset of a training sequence is presumed to be perfectly known. This assumption, however, cannot always be held, and a valid assumption could be that the timing offset is randomly distributed over a number of symbols. Thus, the CIR and timing offset could be jointly estimated.
Other objects, features and advantages will become apparent from the detailed description which follows, when considered in light of the accompanying drawings in which:
Different embodiments will now be described more fully hereinafter with reference to the accompanying drawings, in which preferred embodiments are shown. Many different forms can be set forth and described embodiments should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope to those skilled in the art. Like numbers refer to like elements throughout.
A receiver and associated method estimates a channel impulse response of a communications signal transmitted within a wireless communications system. A communications signal is received as a burst of transmitted symbols including a known training sequence. The timing offset is determined by peak cross-correlation between the received signal and the known training sequence. An initial channel impulse response is estimated based on the cross-correlations at the timing offset. A constant modulus interference removal iteration is applied to improve the initial channel impulse estimation.
In accordance with another aspect, channel parameters are estimated as channel impulse response components that are processed for estimating the initial channel impulse response. A least squares (LS) estimation can be applied in some cases for estimating channel parameters. The timing offset of the known training sequence of the transmitted symbols is randomly distributed over a number of symbols. At least about three iterations are applied and can be implemented using a digital signal processor (DSP). A communications signal is typically formed in accordance with a global system for mobile communications (GSM).
In yet another aspect, a channel estimator circuit and equalizer circuit are used. The equalizer circuit can be formed as a single antenna interference cancellation equalizer circuit, although it should be understood that conventional non-SAIC and SAIC receivers can be used.
As is known to those skilled in the art, the accuracy of the channel impulse response (CIR) estimation typically has a direct impact on the error rate performance of a GSM receiver, as specified by 3GPP, for example, in sensitivity and interference reference performance test cases. A least squares (LS) CIR estimator, however, does not always provide an appropriate margin for some test cases. The described embodiments explained below improve the CIR estimation to ensure that mobile wireless communications devices, for example, handhelds, pass test cases with a significant comfort margin.
In accordance with non-limiting examples of preferred embodiments, to reduce the computation in the case of joint timing and CIR estimation, it is possible to start with a cross-correlation based timing and CIR estimation, and apply the CMIR iterations to the initial CIR estimation from the cross correlations. The penalty for this process was found to be less than ⅕ dB in the SIR/SNR (Signal-to-Interference/Signal-to-Noise) region of interest. Furthermore, the required number of iterations can be substantially reduced by finding an appropriate step size for the iterations. By applying the iterative CMIR (Constant Modulus Interference Removal) to improving the CIR estimation, and leaving any Viterbi equalizer metrics unchanged, the performance in both interference and AWGN (Additive White Gaussian Noise) dominant cases can be improved. For example, the implementation in a 16-bit DSP showed that a 3-iteration CMIR costs 5808 DSP cycles, about 5 MIPs (million instructions per second) for four slots down and 1.25 MIPs for a single slot GSM. It improves the FER/RBERib (Frame Error Rate/Residual Bit Error Rate) performance by 0.5-1.5 dB (decibel) compared with the current LS CIR estimator.
The embodiments as described differentiate prior art solutions in several aspects, including but not limited to:
1) The process and receiver starts with the cross-correlation based timing and CIR estimation, which greatly reduces the computation while minimizing the performance loss. The cross-correlation is done by convolving the full length of the training sequence with the received signal, and a smoothing window, optimized with a length of three symbols, is applied to the cross-correlation to improve the timing estimation. The computation is further reduced by substantially reducing the number of required iterations; and
2) The overall solution is simplified by applying the iterative CMIR to improving CIR estimation and leaving the equalizer structure unchanged. It eliminates the performance loss in Additive White Gaussian Noise (AWGN) cases.
For purposes of illustration, a known series of components and process are described relative to
Referring now to
The channel estimator 12 in the embodiment shown provides the channel impulse response components hi to an equalizer 14, which is a single antenna interference cancellation equalizer. Other kinds of equalizers may be used with the channel estimator 12, or combinations of different kinds of equalizers may be used, with criteria for selecting one or another of the equalizers depending on measurements indicating which of the equalizers would be most suitable under current operating conditions. For example, a receiver might include both a SAIC equalizer and a conventional (non-SAIC) equalizer, and the channel estimator might provide the channel impulse response components hi to a selector module, along with a measurement indicating whether or not a strong interfering signal is present. The selector module could select either the SAIC equalizer or the conventional equalizer, depending on the measurement provided by the channel estimator module. The decision is usually made by comparing the measurement value with a predetermined threshold. A conventional (non-SAIC) equalizer could minimize either the sequence error probability, in what is typically called a Maximum-Likelihood Sequence Estimation (MLSE), or the symbol error probability, in what is typically called a Maximum A Posteriori (MAP) equalizer.
The SAIC equalizer 14 can use either a blind SAIC algorithm, or a joint detection SAIC algorithm. Further details of this algorithm are described in U.S. Patent Application Publication No. 2004/0170234 and article entitled, “Cochannel Interference Suppression for Constant Modulus Signals,” identified above. The SAIC equalizer 14 in this example is followed by a deinterleaver circuit 15 and channel decoder 16 with output symbols.
Referring now to
The SAIC equalizer 14 determines values for a filter in case of a linear SAIC equalizer 14 (block 23), or performs a trellis search in case of a nonlinear equalizer structure. The SAIC equalizer 14 provides the equalizer output (symbols according to, for example, the equalization filter) to the deinterleaver 15, which then provides deinterleaved symbols to the channel decoder 16 (block 24).
As noted before, the channel impulse response (CIR) estimation has an impact on the error rate performance of a mobile GSM receiver. There now follows details of preferred embodiments as shown in the non-limiting examples of
As shown in
As will be explained below, the known training sequence of transmitted symbols is typically randomly distributed over a number of symbols. There now follows a description of basic CIR estimation followed by details of preferred embodiments. The basic algorithms, in accordance with non-limiting examples, will now be described.
It should be understood that the conventional Least Squares (LS) CIR estimation minimizes the following cost function:
In these systems, the cost function for CIR estimation is as follows:
The rationale behind (3) is that when noise is free and the interference has a constant modulus (envelope), the influence of the interference on CIR estimation is removed. Therefore, the cost function (3) is referred to as constant modulus interference removal (CMIR), although in reality the interference in the received signal hardly preserves the constant modulus property due to fading and multiple interferers. The cost function explained relative to (3) improves the CIR estimation for both interference and Additive White Gaussian Noise (AWGN) dominant cases.
Instead of trying to find the closed form solution for the minimization of (3), an iterative numerical solution can use a gradient-based search:
The LS CIR estimation by (1) can be used as an initial value for the iteration, and a number of ten (10) iterations was used in various simulations.
It should be understood that the CMIR iteration can be applied if the timing offset of the burst is exactly known. Due to the time-varying nature of the GSM channel, the timing offset of the burst (and hence the offset of the training sequence) is assumed to be randomly distributed over a number of symbols. A straightforward extension of the CMIR to this case would be a joint timing offset and CIR estimation (assuming that the number of timing offsets is K):
The performance of this LS-CMIR algorithm was evaluated for a half rate speech channel in a typical urban fading condition and a gain of 1.5 dB over the conventional LS was observed in terms of frame erasure rate (FER). A concern is the complexity. Initial complexity estimation showed that about 2,000 DSP cycles would be needed for each of the CMIR iterations. This could be exorbitant for some applications, even with a modest number of timing offsets (e.g. K=7).
To reduce the computation load, the Algorithm can be modified by taking the CMIR iteration out of the timing search loop. The timing search is done jointly by the conventional Least Squares (LS) CIR estimation and the CMIR is applied to the final LS CIR estimation.
While the computation load for CMIR is reduced by K-fold, the performance is degraded noticeably. Simulation for the same channel and fading conditions showed that about half of the gain in the first algorithm. LS-CMIR was lost in second algorithm, LS-CMIR1.
It was observed that the peak of the cross correlations between the known training sequence (TS) and the received samples provides a better timing offset estimation than the LS counterpart. This lead to a modification of the algorithm as indicated:
Compute the cross-correlations (xcs) between full TS and received samples covering the full timing offset range;
Filter the |xcs|2 by a rectangular window with length XCINT;
Find the maximum of the filtered |xcs|2. The corresponding timing offset txc can be calculated and the associated CIR estimation hxc can be extracted from the xcs at the timing offset.
Apply CMIR iterations on hxc by (4)-(9) to give an improved CIR hopt;
This third algorithm (XCF-CMIR1) can be used for joint timing and CIR estimation.
It was noted that XCINT=3 is an adequate choice for the timing offset estimation. A full training sequence, however, rather than a shortened training sequence, could be used. The simulation showed that the XCF-CMIR1 algorithm incurs very little loss comparing to the LS-CMIR algorithm in a low to medium carrier-to-interference (C/I) or signal-to-noise (S/N) region. The cross-correlation with a shortened TS (XCS-CMIR1) suffers loss similar to that of LS-CMIR1.
The number of iterations in the XCF-CMIR algorithm can be substantially reduced without significant loss if a proper step size is used. It is shown in the graphs of
The XCF-CMIR has been implemented in a 16-bit DSP, for example, and costs 5808 DSP cycles for the CMIR with 3 iterations (˜5 MIPS for 4 slots down or 1.5 MIPS for single slot GSM). The performance of the implementation was verified in
The XCF-CMIR also shows improvement in AWGN cases (as specified as reference sensitivity performance by 3GPP). This is shown by
In the simulations, the number of trials conducted was 20,000. The method along with the step size and the iterations (enclosed in parentheses) used in the method are shown as the legend in the figures.
The iterative CMIR can be used to improve the accuracy of the CIR estimation. When timing offset is unknown and needs to be jointly estimated, the three algorithms can be applied as described, for example, the LS-CMIR, LS-CMIR1 and XCF-CMIR1 algorithm. The XCF-CMIR1 algorithm reduces the computations greatly while achieving a similar performance as a least squares CMIR. Simulations showed that the XCF-CMIR1 algorithm improves the error rate performance by ˜0.5 to 1.5 dB in the interested SIR/SNR regions, comparing to the conventional LS CIR estimator.
Many modifications and other embodiments will come to the mind of one skilled in the art having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Therefore, it is understood that the specific embodiments disclosed are not limited, and that modifications and embodiments are intended to be included within the scope of the appended claims.